Kernel based regularisation parameter and source dependent depth weighting in gravity data inversion

In this paper, an inversion algorithm for gravity data is presented. The algorithm exploits a model weighting matrix derived from multiplication of compactness and the modified depth weighting, with its exponent being dependent on the source type. Regularisation parameter and the exponent of depth weighting are the critical inverse parameters: the first is adopted according to the maximum value of the kernel matrice (MaxKer) and the second takes the value of the structural index associated with the source type of interest. The productivity of the proposed method is examined through three synthetic and two real data sets. According to the derived results from synthetic examples, the optimised value found for the regularisation parameter to successfully invert the perfect and noisy data sets are 10-7xMaxKer and 10-1xMaxKer, respectively. Ultimately, we manipulate the inversion algorithm on the two real data sets, from the: i) salt dome of Rogun in Tajikistan and ii) Safo manganese mine in Iran. The suggested value for the regularisation parameter in real cases is 0.05xMaxKer, which is close to the corresponding value to invert the noise contaminated data. Since source types for both cases were known from a priori information, 2 is attributed to the exponent of depth weighting.